In prognostics, the modeling of the failure models is complicated even for a single component, such as fatigue crack growth. For a complex system, there are a large number of components, and hence the failing models can be even more complicated due to diverse sub-systems and their components. The remaining useful life (RUL) of the system, as a whole, depends on many factors and there are often sudden changes in its progression pattern. The interacting multiple model (IMM) estimator is a filtering technique that tracks multiple models and reports the probability of each model. The information fusion ability of IMM with a built-in probabilistic metric is highly desirable in failure model tracking and higher level fusion. A general framework is proposed to describe the system health by a health index, then the RUL can be evaluated as the current health value divided by the degradation rate of the health index at that moment. Within the general framework of a health index, an IMM estimator is proposed to identify the failure models and evaluate both the values and the confidence interval of the RUL. Simulations on various health degradation models are carried out to illustrate the effectiveness of the IMM based RUL estimation. Specifically, the RUL sub-models can be with nearly constant degradation rate, with accelerated growing degradation rate, or some drastic break-down due to environmental changes such as a hard failure. In simulations, the truth is known, and hence the performance of the RUL estimator can be precisely assessed. This paper has not only proposed a fusion scheme to handle various failure models, but also presented the data generation procedure of health index in various situations. Such data sets can be used as benchmarks to compare various prognostics techniques.